{"id":"W3200175561","doi":"10.48550/arxiv.2109.05587","title":"On the Efficiency of Subclass Knowledge Distillation in Classification Tasks","year":2021,"lang":"en","type":"preprint","venue":"arXiv (Cornell University)","topic":"AI in cancer detection","field":"Computer Science","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"University of Toronto","funders":"","keywords":"Computer science; Distillation; Class (philosophy); Machine learning; Task (project management); Artificial intelligence; Binary number; Process (computing); Binary classification; Subclass; Limiting; Annotation; Measure (data warehouse); Data mining; Mathematics; Engineering; Support vector machine","routes":{"ca_aff":true,"ca_fund":false,"ca_venue":false,"about_ca":false,"invisible_to_affiliation_only":false},"retraction":null,"screen":null,"direct_labels":[],"prediction":{"model_version":"metacan-v3-hybrid-931329e0061c","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.003547274,0.001716476,0.001512611,0.001026135,0.0008563094,0.001826805,0.001815364,0.001816361,0.003675078],"category_scores_gemma":[0.01605124,0.0005761448,0.0008880448,0.0009587401,0.001328761,0.004961265,0.003138735,0.003532283,0.001535907],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.000983505,"about_ca_system_score_gemma":0.001911723,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.00668908,"about_ca_topic_score_gemma":0.005847734,"domain_scores_codex":[0.997892,0.0008332726,0.0001393184,0.000536867,0.0003999912,0.0001987242],"domain_scores_gemma":[0.9913689,0.006566438,0.0003403527,0.0009748497,0.0005429573,0.0002065335],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"design_other","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.001050083,0.0004623681,0.004989137,0.0003823674,0.0001448102,0.000179325,0.0003505816,0.2586835,0.01250987,0.01517214,0.003903962,0.7021719],"study_design_scores_gemma":[0.0000347819,0.0001326749,0.0007579922,0.00003623672,0.00003453615,0.00006503533,0.00005279072,0.9816477,0.0047752,0.01110614,0.001339201,0.00001768271],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.1419108,0.004285499,0.8390797,0.001886328,0.0001380634,0.0002342768,0.0004686571,0.003540232,0.008456453],"genre_scores_gemma":[0.7809858,0.001482282,0.2095443,0.0006190366,0.0001534338,0.0002689139,0.001028076,0.0003176286,0.005600624],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.00668908,"threshold_uncertainty_score":0.01875997,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.09450960542890219,"score_gpt":0.2086550890551377,"score_spread":0.1141454836262356,"validation_status":"score_only:v0-immature-baseline","note":"Baseline scores from an immature model (maturity gate not passed). Scores rank; they never assert a category."}}